Earth and Planetary Science Letters 404 (2014) 319–331

Contents lists available at ScienceDirect

Earth and Planetary Science Letters

www.elsevier.com/locate/epsl

Experimental derivation of nepheline syenite and phonolite liquids by partial melting of upper mantle peridotites ∗ Didier Laporte a,b,c, , Sarah Lambart d,1, Pierre Schiano a,b,c, Luisa Ottolini e a Clermont Université, Université Blaise Pascal, Laboratoire Magmas et Volcans, BP 10448, 63000 Clermont-Ferrand, France b CNRS, UMR 6524, LMV, 63038 Clermont-Ferrand, France c IRD, R 163, LMV, F-63038 Clermont-Ferrand, France d Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA 91125, USA e CNR-Istituto di Geoscienze e Georisorse (IGG), Unità di Pavia, Via A. Ferrata 1, 27100 Pavia, Italy a r t i c l e i n f o a b s t r a c t

Article history: Piston-cylinder experiments were performed to characterize the composition of liquids formed at very Received 24 April 2014 low degrees of melting of two fertile lherzolite compositions with 430 ppm and 910 ppm K2O at 1 and Received in revised form 30 July 2014 1.3 GPa. We used the microdike technique (Laporte et al., 2004)to extract the liquid phase from the Accepted 4 August 2014 partially molten peridotite, allowing us to analyze liquid compositions at degrees of melting F down to Available online xxxx 0.9%. At 1.3 GPa, the liquid is in equilibrium with olivine + orthopyroxene + clinopyroxene + spinel in Editor: J. Brodholt all the experiments; at 1 GPa, plagioclase is present in addition to these four mineral phases up to about ◦ Keywords: 5% of melting (T ≈ 1240 C). Important variations of liquid compositions are observed with decreasing upper mantle temperature, including strong increases in SiO2, Na2O, K2O, and Al2O3 concentrations, and decreases lherzolite in MgO, FeO, and CaO concentrations. The most extreme liquid compositions are phonolites with 57% partial melting SiO2, 20–22% Al2O3, Na2O + K2O up to 14%, and concentrations of MgO, FeO, and CaO as low as 2–3%. phonolite Reversal experiments confirm that low-degree melts of a fertile lherzolite have phonolitic compositions, mantle metasomatism and pMELTS calculations show that the amount of phonolite liquid generated at 1.2 GPa increases from potassium 0.3% in a source with 100 ppm K2O to 3% in a source with 2000 ppm K2O. The enrichment in silica and alkalis with decreasing melt fraction is coupled with an increase of the degree of melt polymerization, + which has important consequences for the partitioning of minor and trace elements. Thus Ti4 in our + experiments and, by analogy with Ti4 , other highly charged cations, and rare earth elements become less incompatible near the peridotite solidus. Our study brings a strong support to the hypothesis that phonolitic lavas or their plutonic equivalents (nepheline syenites) may be produced directly by partial melting of upper mantle rock-types at moderate pressures (1–1.5 GPa), especially where large domains of the subcontinental lithospheric mantle has been enriched in potassium by metasomatism. The circulation of low-degree partial melts of peridotites into the upper mantle may be responsible for a special kind of metasomatism characterized by Si- and alkali- enrichment. When they are unable to escape by porous flow, low-degree melts will ultimately be trapped inside neighboring olivine grains and give rise to the silica- and alkali-rich glass inclusions found in peridotite xenoliths. © 2014 Elsevier B.V. All rights reserved.

1. Introduction were focused on partial melting of depleted mantle at mid-ocean ridges (where the mantle contains only 60 ppm K2O according Basaltic magmas at mid-ocean ridges, oceanic islands, intracon- to Workman and Hart, 2005) and thus used starting peridotites tinental rifts and subduction zones are mainly produced by partial either K2O-free (e.g., Falloon et al., 1999, 2008; Robinson et al., melting of upper mantle peridotites. Many experimental studies 1998; Presnall et al., 2002) or very poor in K2O (e.g., Baker and Stolper, 1994). Estimates of K2O in the primitive mantle (160 to 340 ppm; Taylor and McLennan, 1985; McDonough and Sun, 1995; * Corresponding author at: Laboratoire Magmas et Volcans, CNRS/UBP/IRD, 5 rue Arevalo et al., 2009, and references cited therein) are, however, up Kessler, 63038 Clermont-Ferrand Cedex, France. Tel.: +33 4 73 34 67 33. to 6 times higher than in the depleted MORB mantle, and there E-mail address: [email protected] (D. Laporte). 1 Now at: Lamont Doherty Earth Observatory, Columbia University, Palisades, NY are source regions in the upper mantle that may be even more 10964, USA. enriched (McKenzie, 1989). In particular, the lithospheric mantle http://dx.doi.org/10.1016/j.epsl.2014.08.002 0012-821X/© 2014 Elsevier B.V. All rights reserved. 320 D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 below continental and oceanic crust may be enriched in potassium in the microdike is not in close contact with peridotite minerals, it by repeated episodes of metasomatism: for instance, Menzies et al. is not modified by the growth of quench crystals at the end of (1987) describe metasomatized peridotite xenoliths in which K2O the experiment. In all our runs except MBK+7 and MBK15, we exceeds 1%. observed a few microdikes varying from tens to hundreds of mi- As K2O is highly incompatible with respect to most minerals crons in length and from a few microns to a few tens of microns during mantle melting, and as the incorporation of alkalis strongly in width, and yielding consistent glass compositions (Table 1). The affects the degree of polymerization of silicate liquids (Ryerson, length of the microdikes is short compared with the thickness of 1985; Hirschmann et al., 1998), even a moderate increase of K2O the graphite walls (0.7 to 0.9 mm) so that the partial melt is not in in the mantle source may have a dramatic effect on the major el- contact with the outer platinum container. This technique is per- ement composition and on the structure of partial melts, specially fectly suited to the study of low degrees of melting because the at low degrees of melting (≤5%; all percentages are wt.% except volume of the microdikes is very small in comparison with the to- when specified otherwise). These changes in melt structure may tal volume of partial melt in the sample. in turn have important consequences for the partitioning of minor A drawback of the experiments at low degrees of melting is that and trace elements (Gaetani, 2004; Huang et al., 2006). any trace of water in the capsule will partition preferentially into Here we present the results of melting experiments on two the partial melt, yielding significant amounts of dissolved water fertile, K2O-rich peridotites: composition MBK (410 ppm K2O) is even under nominally anhydrous conditions. Dissolved water con- within the range of the primitive mantle as estimated by Arevalo tents in some experimental glasses were measured using an ion et al. (2009): 340 ± 140 ppm K2O; composition MBK+ is signifi- microprobe with a beam diameter of 5μm (see Section 1 of the cantly more enriched (930 ppm K2O). Three series of experiments Supplementary material; Ottolini and Hawthorne, 2001; Ottolini et were performed: two with MBK at 1 and 1.3 GPa, and one with al., 1995, 2002; Scordari et al., 2010). The water contents are rela- MBK+ at 1.3 GPa. The pressures of 1 and 1.3 GPa were chosen tively low and decrease with increasing melt fraction F : from 1.1% because the effect of alkalis on the structure of mantle melts is at our lowest degrees of melting (F ≈ 1%) to 0.1% at F = 12.7%. more important below 1.5 GPa (Hirschmann et al., 1998). In ad- dition, this range of pressure straddles the transition from pla- 3. Experimental results gioclase + spinel lherzolite (at 1GPa) to spinel lherzolite (at 3.1. Experiments at 1.3 GPa 1.3 GPa), allowing to study the effect of plagioclase on low-degree melt compositions. We used the microdike technique (Laporte et Run information (P –T conditions, run products, modes, liquid al., 2004)to extract and analyze liquids at degrees of melting F compositions, etc.) is summarized in Table 1; compositions of solid down to 0.9%. Our low-degree melts are rich in Na O + K O (up 2 2 phases are given in Supplementary Table S1. Liquid is in equi- to 14 %), SiO (up to 57%), Al O (20–22%), poor in MgO, FeO, 2 2 3 librium with olivine (ol) + orthopyroxene (opx) + clinopyroxene CaO (down to 2–3%), and classify as trachyandesites, tephriphono- (cpx) + spinel (spl) in all the experiments at 1.3 GPa (Fig. 1b). The lites and phonolites in the total alkali–silica (TAS) diagram (Le ◦ ◦ melt fraction F increases from 0.9% at 1180 C to 4.1% at 1270 C Bas et al., 1986). Calculations with the pMELTS software (Ghiorso ◦ ◦ in MBK, and from 1.6% at 1150 C to 5% at 1250 C in MBK+ (Ta- et al., 2002) confirm the experimental findings and document ble 1). Thus, at a given temperature, F is higher in the composition the effect of bulk K2O (up to 2000 ppm) on melt compositions with the larger K2O content. The absence of microdikes in sample and melt fractions. Lherzolites relatively rich in potassium (HK-66, ◦ MBK+7 (T = 1130 C) and its lower degree of textural equilibra- 0.07% K O; PHN1611, 0.14% K O) were studied by Kushiro and co- 2 2 tion suggest that it could be subsolidus. In this case, a K-bearing workers (Takahashi and Kushiro, 1983; Hirose and Kushiro, 1993; solid phase, such as feldspar, must be present even if we did not Kushiro, 1996), but most of their experiments were at high degrees find it despite careful examination. Alternatively, a quantity of K- of melting and none produced a partial melt as enriched in alkalis rich glass (a few 0.1%) too small to be detected with the scanning as in our work. electron microscope may be present in MBK+7. Oxide concentrations in liquids are plotted as a function of the 2. Experimental and analytical techniques degree of melting in Fig. 2. In both MBK and MBK+, important variations of liquid compositions are observed with decreasing F , Techniques are described in detail in Section 1 of the Supple- including increases in SiO2, Na2O, K2O, and Al2O3, and decreases mentary material. The starting material used to prepare composi- in MgO, FeO, and CaO (the case of TiO is discussed later; Ta- + 2 tions MBK (410 ppm K2O) and MBK (930 ppm K2O) was a spinel ble 1). These compositional trends are qualitatively similar to those lherzolite xenolith (Bri3) from Mont Briançon volcano, French Mas- previously described in the literature (Baker and Stolper, 1994; sif Central. As Bri3 contains only 100 ppm K2O, small amounts of Kushiro, 1996; Pickering-Witter and Johnston, 2000), but the de- + synthetic basalt B2 were mixed to Bri3 to produce MBK and MBK grees of enrichment or depletion are more pronounced: for in- + (compositions Bri3, B2, MBK and MBK are given in Table 1). Ex- stance, near-solidus melts from the literature at 1 to 1.5 GPa con- + cept for their larger K2O contents, MBK and MBK are broadly tain ≈6% of MgO and CaO (Hirschmann et al., 1998; Robinson et similar to fertile peridotites studied in the past (MM3, Baker and al., 1998)or more (Falloon et al., 2008) while our near-solidus Stolper, 1994; MPY, Robinson and Wood, 1998; Table 1). melts at 1.3 GPa contain only ≈3% of these oxides (Fig. 3). Our The experiments were made in a piston-cylinder apparatus, us- most extreme liquid composition (MBK+6) has ≈57% SiO2, 21% ing double containers made of a graphite crucible fitted into a Al2O3, Na2O + K2O ≈ 13.5%, MgO, FeO, and CaO concentrations platinum capsule. Despite the fast quench rate of this apparatus ◦ as low as 2–3%, and plots in the phonolite field in the TAS classi- (≈100 C/s), quench crystals grow at the end of the experiments fication diagram (Fig. 4). Despite the enrichment in SiO2 and the and modify substantially the composition of interstitial liquid in strong depletions in MgO and FeO, it coexists with typical mantle partially molten peridotites (e.g., Baker and Stolper, 1994). In this olivine and pyroxenes (mg# ≈ 89–90), and it is strongly nepheline study, we used the microdike technique (Laporte et al., 2004; Lam- normative (20.5%; Table 1). bart et al., 2009, 2013) to separate small volumes of liquid in equilibrium with mineral phases into fractures of the graphite con- 3.2. Experiments at 1 GPa tainer (Fig. 1a). Microfractures develop into the graphite container at the beginning of the experiment and pump small volumes of In the experiments with MBK at 1GPa, F increases from 0.9% ◦ ◦ ◦ partial melt out of the neighboring peridotite: as most of the liquid at 1190 C to 22.8% at 1310 C; the sample run at 1180 C shows D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 321

al. 14 the

) ) ) ) ) )

+ ) ) ) ) ) ) ) ) ) ) ) ) ) et k 8 30 7 15 9 6 6 6 13 7 21 15 8 6 6 6 5 6 14

in ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( ( 2 Nancy;

network MBK 25 41 40 26 71 32 44 41 06 50 80 64 54 43 43 37 34 41 12

TiO ......

D as percentages

given

and

or

(CRPG,

j

Wasylenki are 13 Volcans.

+ of weight et 776 0 937 133 1 395 0 806 0 140 1 237361 0 0 328 0 451601 0 0 183278 0 316 0 404 0 439 0 0 567608 0 0 546 0 153 1

......

5 NBO/T former

MBK the

.

Minéraux 3 i

); Table SO

et 00 00 90 10 00 20 50 90 00 70 00 50 90 90 80 10 50 50 50 0 0 9 0 0 0 0 40 Magmas

Runs ......

7 uncertainties network 2004 Neph 19

Table S1.

from a

0.22% the MBK.

al., is

as Roches

h

g

et and

and 00 00 028 00 88 66 05 910 022 09 08 221 015 020 016 014 014 011 09 0 2

Laboratoire des ......

Felds either at CO

content

Ti O

Laporte phases 2 60 10 00 40 12 21 71 80 4traces 78 40 30 60 50 30 10 30 20 00 80 00 00 60 Supplementary ( composition ...... g

K 0.09%

in

Sp all ,

B2

ICP-AES d’Analyses

−

the bulk

of

O

gel 2 g of 40 00 80 20 62 92 81 11 71 41 12 52 21 11 02 42 12 72 52 31 33 03 92

...... H and

considering

Cpx Service by

basalt NiO

0.57%

uncertainties g , analyzed by the 87 70 06 04 413 513 513 916 213 418 515 015 615 216 418 118 718 516 216 815 714 017 719

...... +

σ composition O with at

Opx 2 0.23% − experiments

H 1 obtained

6was

+ 518 214 10 617 28 312 319 320 418 319 320 320 220 320 621 322 419 621 620 120 826 030 822 mean

...... g the

(Bri3), 0 5

Ol 2.12% gel ICP-AES

melting

on includes NBO/T

The

of C. 20 760 862 861 961 061 460 951 747 560 958 759 559 159 656 255 657 156 757 057 054 g

...... ◦ Central total 0

based partial Liq

includes

Phonolite (1977) . the 1150

values are

are

also analyzed by Massif f

393 812 222 216 30 71 23 95 80 31 73 64 314 521 71 92 93 93 55 32 50 65 60 50 22 55 8 10

......

and and 20%). (the

MM3, Dun2 ±

Provost mg#

total

were GPa

For French cited and

1

and

cation B2 e

error:

09 60 50 74 84 76 73 74 1550 70 27 71 17 72 72 94 68 0871 69 45 71 51 72 72 17 68 5777 69 29 70 25 71 50 71 33 70 72 14 74 71 89 00 89 44 69 00 89 00 90 41 51 62 90

at ...... unit

gel Total

analytical

run MBK21,

100%). least Briançon,

and Albarède

3

to The (relative

to respectively. O 09 98 60 98 39 100 54 94 6666 92 50 95 66 98 97 15 90 7466 97 98 5949 98 99 33 96 2394 94 52 95 01 96 73 96 94 97 97 63 99 87 98 21 99 08 100 44 100 98 100 61 99 37 99 was

...... the 2 coordinated

Bri3 Al from Mont of

1981 ).

MBK14 which

2

from 86 17 64 14 29 22 07 16 7314 19 39 19 08 19 18 70 21 5065 18 18 5053 21 21 03 20 8997 21 66 20 12 20 74 20 51 19 18 28 18 82 3 8899 4 4 1150 21 3 04 20 70 4 6, ...... terms

SiO Price, + spectroscopy recalculated

56 modified in

peridotite MBK4,

Wood (1998) ,

tetrahedrally

gel and 3 to of

been O

. and per 2 given 16 49 37 50 00 56 17 50 05 55 0203 53 51 01 54 0003 57 57 0204 55 04 54 04 51 50 0201 56 02 54 04 54 02 53 04 52 51 04 49 49 44 4847 44 44 0368 56 45 45 44

...... peridotite +

0 2 program Irving are

MBK1 have Fe absorption (

phonolite as spinel

oxygens 17 0 16 0 05 0 15 0 0506 0 13 0 13 0 0 0602 0 0 0507 0 12 0 12 0 0 0408 0 07 0 08 0 11 0 09 0 0 12 0 13 0 1313 0 0 0413 0 0 12

......

is Runs table Robinson of

.

balance compositions atomic

+

the Germany iron and

in The 57 0 42 0 97 0 01 0 7310 0 67 0 72 0 0 7900 0 0 7097 0 51 0 90 0 0 1016 0 52 0 22 0 79 0 43 0 0 94 0 04 0 0196 0 0 1018 0 0 92 0 55 mass parentheses)

MBK all ......

FeO MnO Cr powdered a

(in a

and proportions.

non-bridging experiment

2 Heldburg, analyzed by using of 73 6 47 7 52 1 69 7 0204 3 91 4 88 4 5 6153 1 2 6786 2 93 3 02 5 5 6070 2 78 3 87 3 91 4 89 4 5 50 5 13 8 1415 8 7 51 2 11 7 33 1 17 7

......

phonolite errors

phase mixing was

Stolper (1994) respectively.

from

composition

fractions , The

by

and + 61 0 28 0 76 0 79 0 7187 1 74 1 81 0 0 4138 0 40 0 97 0 1 9932 0 0 8854 0 06 0 22 0 89 0 16 0 0 10 0 13 0 2648 0 0 95 0 57 0 42 0 38 0

number Bri3 ...... and

3 bulk calculated

Heldburg in

crystallization

melt. the MBK cation of

O

a melts, 2 lherzolite

Baker

and were K

and are is

produced and

of OCaOTiO ); 13.

77 1 37 11 19 11 162324 5 83 6 7 9 27 11 756336 3 15 5 8 8 7819 1 2 256811 2 44 4 95 5 18 6 6 8 01 3 14 12 28 2 041 3 64 1 093 3 008 3 +

2 ...... 2 NBO/T, + cpx MPY

spinel partial

Fe

of MBK were

of

MBK products

experiments

+ in and

68 6 94 0 36 0 9682 3 81 2 63 1 0 30 0 5054 4 23 2 75 1 1 1743 6 6 7233 6 90 4 05 4 55 3 01 2 2 05 0 75 2 65 6 63 0 60 16 4 91 0 30 0 Cryspho2 ......

except

+

run 2 between

MBK 2 Run

compositions sample Mg

the

xenoliths melting

calculation 6.

(wt.%). TiO experimental in

OMgOK and

in 57 1 9308 4 59 5 23 6 49 8 10 52 13 04 3 1874 8 8 5046 2 2 16 5 00 11 2758 2 35 4 95 4 31 6 74 6 8 50 9 33 39 12 2 17 1 3538 38 37 3140 38 38

+ 2

......

96.13:3.87 of 0 0 0 0

the

10 Na

and where mantle

MBK MBK partial in compositions ), found

and

carrying content +

phases

2

are run (all

of Fe C) was

fertile

7

◦ of T ( coefficient + + materials

account different).

the

+

totals grain 2

phonolite

98.58:1.42

nepheline MBK

are a

into compositions

to /(Mg 0 1150 8 000 11800 11900 very low 1200 melt fraction,0 not analyzed 1220 5 1240 6 0 1260 5 4 1310 2 1 33 1180 12003 7 3 7 3 11803 1200 1250 8 1270 7 5 4 0 1290 2 333 11503 11803 1200 7 3 1220 6 1240 6 1250 5 5 4 3 1130 no liquid detected

of starting ...... are partition slightly

1 +

percentages

P Table S1.

1 (GPa) 2

MPY + of feldspar

taken

B2

experiments analytical the

just Mg

mantle

c

is normative and MBK d = one

not

weight 2

and are 7 61 51 41 31 21 11 1413 1 1 a

68

b is . + + + + + + + + + + a b

TiO + D reversal Fertile MM3 Composition Runs Original mg# The Only CIPW Ti Bri3

i j f

c gel MM3 MBK BRI3 0 Heldburg MBK MBK gel B2 1 MPY MBK MBK MBK MBK MBK MBK MBK MBK CrysPho2 1 MBK4MBK3MBK2 1 MBK1 1 MBK15 1 1 MBK19 1 MBK16MBK18 1 MBK20 1 Dun2 1 MBK21 1 1 1 MBK14 1 a e g k b d h (2003) Table 1 Compositions of http://helium.crpg.cnrs-nancy.fr/SARM/pages/roches.html are Supplementary breaker 322 D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331

◦ Fig. 1. (a) SEM-Backscattered electron microphotograph of sample MBK3 (1.3 GPa-1200 C) showing a microdike filled with quenched melt (“glass”) into the graphite container (black). (b) Texture and paragenesis of partially molten peridotite MBK3; solid phases are as follows: Ol is grey, Opx is slightly darker than Ol, Cpx is light grey, and Spl is ◦ white. (c) Texture and paragenesis of partially molten peridotite MBK14 (1 GPa-1190 C): pl (dark grey) is present in addition to ol, opx, cpx, and spl. The interstitial glass is barely visible in b and c due to its low fraction (1.7% in b; 0.9% in c); the dark voids correspond to grains plucked off during sample preparation. Scale bars: 60 μm in a; 20 μm in b and c. traces of liquid, but we could not get reproducible glass analyses in ies can readily be explained by differences in bulk compositions: ◦ this case. In a plot of F vs. T , the experiments run at T ≤ 1220 C for instance, the lower maximum CaO content of MBK melts (at define a linear trend that intersects the solidus (F = 0%) at T = the cpx-out temperature; Figs. 2–3) is related to the lower modal ◦ 1178 C. The main difference with the experiments at 1.3 GPa is abundance of cpx in MBK (Table 1), and the lower Na2O + K2O to- the presence of up to 3% plagioclase (pl) at low F (Fig. 1c), in addi- tals in MM3 melts (Fig. 4) are due to the lower bulk K2O content tion to ol, opx, cpx, spl, and liquid. The pl composition ranges from of MM3 (<100 ppm; Table 1). ◦ ◦ An54Ab43Or3 at 1180–1190 C to An57Ab41Or2 at 1220 C. Plagio- ◦ clase disappears slightly below 1240 C, at F ≈ 5%. Accordingly, 3.3. Reversal experiments despite its low fraction in the subsolidus state (3%), pl is stable ◦ ◦ up to ≈60 C above the solidus. From 1240 to 1290 C, the liquid We made a series of runs to reverse partial melting experiment + ◦ is in equilibrium with ol + opx + cpx + spl. Over that T range, MBK 6 at 1.3 GPa-1150 C. This experiment was chosen because it has the most extreme liquid composition: the aim was to verify cpx is the main reactant and ol is on the product side of the melt- that a phonolite composition can be at chemical equilibrium with ing reaction: 0.78 cpx + 0.21 opx + 0.12 spl → 1.0 liq + 0.11 ol a lherzolite. The starting material in the reversal experiments was (see Section 2 of the Supplementary material). Clinopyroxene dis- ◦ ◦ made of peridotite MBK+ and phonolite gel +6, which is equiva- appears between 1290 and 1310 C, at F ≈ 22%. At 1310 C, the lent to the glass in run MBK+6 (except that it is ≈10% richer in liquid is in equilibrium with ol, opx, and traces of Cr-rich spl. Na O; Table 1). We first used a sandwich configuration with sep- Liquid compositions in MBK at 1 GPa show strong increases 2 arate layers of phonolite and peridotite, but then we switched to in SiO , Na O, K O, and Al O , and decreases in MgO, FeO, and 2 2 2 2 3 a homogeneous mixture of peridotite and phonolite because the CaO with decreasing F (Fig. 2; Table 1). Only the liquids produced initial chemical heterogeneity in the sandwich configuration may at high F (13 to 23%) plot in the basalt field in Fig. 4: the liq- impede the approach to equilibrium. So we equilibrated a mixture uids produced at F ≤ 5.4% are nepheline normative and plot in the of 80.2% MBK+ and 19.8% gel +6 at P = 1.3 GPa and T = 1180 ◦ trachybasalt to trachyandesite fields. In detail, comparison of melt and 1200 C (MBK+14 and MBK+13, respectively; Table 1), and compositions from MBK at 1 and 1.3 GPa indicates that the pres- we used the microdike technique to analyze glass compositions. ≈ ◦ ence of pl has a distinct effect on Na2O (buffered at 6%; Fig. 2), Experiments MBK+14 and MBK+13 were run 30 to 50 C above ◦ Al2O3 (buffered at 19.5%), and K2O. At a given F , K2O is smaller at the temperature of MBK+6 (1150 C) because the liquid phase 1 GPa because it is less incompatible in the presence of pl, and CaO was expected to contain less dissolved water in experiments with is slightly larger. As a consequence of the effect of pl, experiments F ≈ 20% than in a low-degree melting experiment with F = 1.6% = MBK4 and MBK14, which have the same melt fraction (F 0.9%), (see also Laporte et al., 2004). For instance, partial melts in runs + have very different values of Na2O K2O: 9.1% in MBK14 (1 GPa) MBK3 (F = 1.7%) and MBK20 (F = 12.7%) contain 1.09% and 0.14% vs. 12.8% in MBK4 (1.3 GPa). H2O, respectively. This is due to the highly incompatible behavior The trend of liquid compositions from MBK14 to MBK21 is close of water: for a given degree of water contamination in the sample to that reported for fertile peridotite MM3 at 1 GPa by Baker and capsule, the higher the melt fraction, the lower the amount of wa- Stolper (1994), Baker et al. (1995) and Hirschmann et al. (1998), ter dissolved in melt. The higher temperature in reversals MBK+13 except for the effects of pl (not found in these studies). In par- and MBK+14 in comparison to MBK+6 is aimed at balancing the ticular, both trends culminate at 55.7% silica close to the solidus lower water content in the melts and its impact on the extent of (Fig. 2), and our lowest degree melt (F = 0.9%) is just slightly melting. ◦ more depleted in MgO, FeO, and CaO than their lowest degree In the reversal experiment at 1180 C, the melt fraction (≈15%) melt (F = 2.3%). Differences between the two groups of stud- is lower than the percentage of gel +6in the starting mixture D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 323

Fig. 2. Plots of major oxide concentrations in experimental partial melts as a function of the degree of melting (F , wt.%) for fertile peridotites MBK and MBK+ at 1 to 1.3 GPa. Symbols are as follows: open green squares for MBK at 1 GPa; solid red circles for MBK at 1.3 GPa; and open red circles for MBK+ at 1.3 GPa (colors on the Web only). The results of program pMELTS are shown by the blue curves: the dark blue curve is for MBK+ at 1.3 GPa (for F ≤ 5% as in the experimental series); the light blue curve is for MBK at 1 GPa (for F ≥ 5%, that is, out of the experimental pl stability field). The small black crosses are the data for fertile peridotite MM3 at 1 GPa (Baker and Stolper, 1994; Hirschmann et al., 1998). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) 324 D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331

Fig. 3. Composition of partial melts of fertile peridotites between 1 and 1.5 GPa Fig. 4. Composition of partial melts of fertile peridotites between 1 and 1.5 GPa plotted in the CaO–MgO diagram. Our partial melting experiments at 1.3 GPa are plotted in the total alkali-silica diagram (Le Bas et al., 1986). Our partial melting shown by the red circles (colors on the WEB only): open circles for lherzolite experiments at 1.3 GPa are shown by the red circles (open circles: MBK+; solid cir- + MBK ; solid circles for lherzolite MBK. The open green squares are for the ex- cles: MBK); the open green squares are for the experiments with MBK at 1 GPa periments with MBK at 1 GPa. The open red circle with a central dot is reversal (colors on the Web only). The open red circle with a central dot is reversal ex- + + + experiment MBK 13: partial melting experiment MBK 6 and its reversal MBK 13 periment MBK+13: partial melting experiment MBK+6 and its reversal MBK+13 are surrounded by the red dashed line. The small black crosses correspond to fer- are surrounded by the red dashed line. The small black crosses correspond to fertile tile peridotite MM3 at 1 GPa (Baker and Stolper, 1994; Hirschmann et al., 1998). peridotite MM3 at 1GPa (Baker and Stolper, 1994; Hirschmann et al., 1998); the di- The other near-solidus melts from the literature are plotted for comparison: the di- amonds correspond to a K2O-free analogue of MM3 at 1 and 1.5 GPa (open and solid amonds correspond to a K2O-free analogue of MM3 at 1 and 1.5 GPa (open and diamonds, respectively; Falloon et al., 2008). The solid triangle is the near-solidus solid diamonds, respectively; Falloon et al., 2008); the solid triangle is the near- melt of MORB-pyrolite MPY at 1.5 GPa (Robinson et al., 1998). The blue symbols cor- solidus melt of MORB-pyrolite MPY at 1.5 GPa (Robinson et al., 1998). The blue respond to Heldburg phonolite (Irving and Price, 1981)and to the glass analyzed in symbols correspond to a phonolite carrying spinel lherzolite xenoliths (Heldburg the crystallization experiment CrysPho2 (crystallization of a composition equivalent ◦ phonolite; Irving and Price, 1981) and to the glass analyzed in the crystallization to partial melt MBK+6 at 1 GPa-1150 C). The star labeled “DGA-40” corresponds to ◦ experiment CrysPho2 (crystallization of a composition equivalent to partial melt a trachyte melt in equilibrium with ol + opx + cpx at 1.2 GPa and 1150 C (Draper + ◦ MBK 6 at 1 GPa-1150 C; Table 1). (For interpretation of the references to color and Green, 1999). Melt compositions computed by pMELTS for MBK+ at 1.3 GPa in this figure legend, the reader is referred to the web version of this article.) are shown by the blue curve (the theoretical trend is shifted to the left of the ex- perimental data because of a slight underestimation of SiO2 by pMELTS). The grid (19.8%), and ≈8% of plagioclase (K-oligoclase) is present in addi- corresponds to the chemical classification of volcanic rocks, with the main fields tion to ol, opx, cpx, and spl (Supplementary Fig. S1a). Accordingly, of interest labeled as follows: (0) Phonotephrite; (1) Tephriphonolite; (2) Phono- + ◦ lite; (3) Trachyte; (4) Trachyandesite; (5) Basaltic trachyandesite; (6) Trachybasalt; the degree of crystallization of gel 6 is relatively large at 1180 C. (7) Basalt; (8) Basaltic andesite. (For interpretation of the references to color in this ◦ At 1200 C (MBK+13), the melt fraction is equal to 21.7% and we figure legend, the reader is referred to the web version of this article.) found only one plagioclase left in a microdike (Fig. S1b–c). Sub- tracting a percentage of melt equal to the percentage of gel +6 product. There is also a good correspondence between the com- and normalizing to 100%, we obtain a mode composed of 59.0% ol, positions of mineral phases in experiments MBK+6 and MBK+13 16.1% opx, 20.8% cpx, 1.7% spl, and 2.3% liq. Accordingly the degree (see Supplementary Table S1), specially for Mg-numbers. The cor- of melting of peridotite MBK+ in MBK+13 is equal within error to respondence is not so good for CaO in cpx, which is lower in the degree of melting in MBK+6 (1.6%). The persistence of traces of MBK+13 compared with MBK+6, and CaO in opx, which is higher. plagioclase (K-andesine) in MBK+13 is presumably due to the fact These discrepancies are presumably a direct consequence of the that gel +6 is 10% richer in Na2O than MBK+6 liquid, thus favor- higher temperature of run MBK+13. Despite some minor differ- ing plagioclase stability at slightly higher temperatures or degrees ences between MBK+6 and MBK+13, we conclude that phase of melting. As in all our reversal experiments, the opx percentage compositions in partial melting experiment MBK+6 are very close in MBK+13 (16.1% after subtraction of gel +6) is lower than in to chemical equilibrium, and that partial melting of a fertile spinel MBK+6 (21.4%), and this goes along with a higher ol percentage: lherzolite produces phonolitic liquids at low degrees of melting. 59.0 vs. 56.6%. These observations are qualitatively in agreement with a reaction relation between opx and liq to produce pl + 4. Theoretical results ol, which is the reaction expected upon cooling of plagioclase + spinel lherzolite (Walter et al., 1995). 4.1. The composition of low-degree melts of MBK+ at 1.3 GPa The liquid in MBK+13 matches well the partial melt composi- tion in MBK+6 except that it is slightly depleted in MgO and CaO We used the software pMELTS (Ghiorso et al., 2002; Smith and (Table 1; Figs. 3–4). As the concentrations of these two elements Asimow, 2005)to verify that phonolite melts can be produced ◦ ◦ increase strongly with T increasing from 1180 C to 1200 C, a re- at low degrees of melting of a fertile peridotite (see Section 3 ◦ versal experiment run just slightly above 1200 C should yield a of the Supplementary material for details). A comparison of the good match for these two elements too, as well as a pl-free run theoretical results with our experimental series at 1 GPa shows D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 325

Fig. 5. Evolution of the modal proportions of solid phases with increasing degree of melting in peridotite MBK at 1 GPa (a) and in peridotite MBK+ at 1.3 GPa (b). The theoretical modes (pMELTS) are shown by continuous lines (the vertical dashed lines correspond to pl disappearance in pMELTS calculations). The symbols for the experimental data are as follows: × for ol; + for opx; circles for cpx; squares for spl (not shown in a for clarity); diamonds for pl. In a, the short, thick line in the lower left corner (intersecting the F axis at 0.7%) corresponds to the pl modes computed by pMELTS. that pMELTS predicts quite accurately the oxide concentrations in tative of the melting behavior at moderate pressures (1–1.5 GPa). liquids (Fig. 2) and the modal proportions of solid phases at de- At P = 1GPaand below, Na2O and K2O become less incompati- grees of melting ≥5% (Fig. 5a). For F < 5%, however, the theoretical ble due to the abundance of pl, and the enrichment in alkalis at results are not in good agreement with the experimental observa- low F is less marked. With P increasing above 1.5 GPa, the effect tions because pl seems to be strongly underestimated by pMELTS, of alkalis on the structure and silica content of mantle melts de- with important consequences for pl mode at the solidus, the melt creases (Hirschmann et al., 1998). Calculations for a lherzolite with fraction at which pl disappears, and pl composition. The calcula- 1000 ppm K2O confirm that the silica content of low-degree melts tions with composition MBK+ at 1.3 GPa are also affected by this (with F ≈ 1.6% and Na2O + K2O = 14%) decreases rapidly with in- problem, but to a lesser extent: only the data at F < 1.6% are con- creasing pressure: from 56.8% at 1.2 GPa, to 54.4% at 1.5 GPa, and cerned. to 50.5% at 2 GPa (Supplementary Fig. S2a). At a given degree of melting (with F ≥ 1.6%), the proportions of The K2O content of the lherzolite source has a strong effect cpx and spl computed for MBK+ at 1.3 GPa are in good agreement on the SiO2 content in melt at a given melt fraction (Fig. 6): at with the experimental values while opx is slightly overestimated F = 3%, for instance, SiO2 in melt increases from 49.3% to 53.8%, and ol slightly underestimated (Fig. 5b). The trends of liquid com- and to 57.0% for bulk K2O contents of 0, 1000, and 2000 ppm, re- positions as melting progresses are very close to the experimen- spectively. The SiO2 increase is accompanied by an increase of the tal trends (Fig. 2). The main systematic differences are for MgO, K2O content in melt (at nearly constant Na2O) and by strong de- which is overestimated by 2.1% in average, and to a lesser extent creases in MgO, FeO, and CaO. A major effect of increasing the K2O for SiO2, which is underestimated by 1.1% in average. These dis- content of the source is to multiply the amount of phonolite liq- crepancies are presumably related to the overestimation of opx uid generated: considering liquids with a total alkali of 14% as a proportion and to the underestimation of ol proportion. Leaving basis for comparison, the amount of phonolitic liquid produced in- aside the case of MgO, the model accurately predicts the evolution creases from 0.27% in a source with 100 ppm K2O to 2.58% in a of melt compositions with increasing degree of melting of MBK+ source with 2000 ppm K2O (Fig. 6). This increase of the fraction at 1.3 GPa. In particular, the liquids calculated at the beginning of phonolite melt goes along with an increase of its SiO2 content of melting (F = 1.6–2%) are silica-enriched (55–56%), phonolitic (from 55.1% to 57.7%) and a strong increase of the ratio K2O/Na2O (Fig. 4), and very poor in FeO and CaO (2–3%; Fig. 2), thus match- (from 0.36 to 1.23; Supplementary Table S2). ing very well the compositions of our lowest degrees of melting (MBK+6, MBK+13). 5. Discussion

4.2. Effect of bulk K2O on the mass fraction and composition of 5.1. Composition of low-degree mantle melts at P ≤ 1.5 GPa low-degree melts Our main result is that low-degree melts of a fertile lherzolite We used pMELTS to evaluate the effect of the K2O content of are strongly enriched in alkalis and silica, reaching the fields of a lherzolite source on the production of silica- and alkali-enriched trachyandesites, tephriphonolites or even phonolites in the TAS di- low-degree melts. The calculations were made on a series of bulk agram (Fig. 4). The best example is provided by our series with compositions (based on Bri3 in Table 1) and containing from 0 composition MBK+ (930 ppm K2O) at 1.3 GPa, but the produc- to 2000 ppm K2O (see the spreadsheet “Effect of K2O” in Supple- tion of phonolites is not restricted to mantle sources so enriched mentary Table S2). We did not take into consideration the results in K2O: a phonolite liquid was produced in our series with compo- at very low F due to the difficulties related to pl stabilization in sition MBK (410 ppm K2O) at 1.3 GPa, and phonolite compositions pMELTS as detailed in the Supplementary material. Thus the dis- are found by pMELTS for a source with 100 ppm K2O. The main cussion hereafter is restricted to liquids with Na2O + K2O ≤ 14%, difference is that a source with 100 ppm K2O will produce only but our general conclusions are not sensitive to the exact limit 0.3% of phonolite melt and ≈1% of tephriphonolite melt as com- chosen. The calculations were made at 1.2 GPa and are represen- pared to more than 3% of phonolite melt and up to 5.2% of tra- 326 D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331

≈6% of MgO and CaO, and 5% of FeO (Hirschmann et al., 1998; Robinson et al., 1998)or more (Falloon et al., 2008). Note that all the liquids in our study, even the most depleted in MgO, are in equilibrium with a spl (±pl) lherzolite paragenesis with typical mantle mg# values (e.g. olivine with mg# ≈ 90).

5.2. Implications for minor and trace element partitioning

The evolution from basalts and trachybasalts to phonolites with decreasing F results in changes in melt polymerization, which may have important consequences for the partitioning of mi- nor and trace elements. The degree of polymerization increases with decreasing F , as indicated by the variations of the ratio of non-bridging oxygens to tetrahedrally coordinated cation (NBO/T; Mysen and Richet, 2005): NBO/T decreases from 0.78–0.94 in basalts (e.g., samples MBK21 and MBK20; Table 1) to 0.33–0.36 in trachyandesites (MBK14 and MBK3), and to 0.18 in phonolites (MBK+6). We did not measure the trace elements in our run prod-

Fig. 6. SiO2 in melts as a function of melt fraction, F , for a lherzolite source with ucts, but the TiO2 data can be used to discuss the interactions K2O equal to 0, 100, 300, 500, 1000, 1500, and 2000 ppm (thick continuous lines) between melt structure and partitioning. In partial melts from at 1.2 GPa; the calculations were made with the software pMELTS. For a given value composition MBK at 1GPa, TiO2 decreases from 1.0 to 0.5% with of bulk K2O, the total of alkalis in the melt increases from right to left (that is, with decreasing F ). Each curve is terminated on the left by a small solid circle increasing F (Fig. 2). On the contrary, TiO2 increases with increas- that corresponds to the point where Na2O + K2O in the melt reaches 14%: the ing F in the two series at 1.3 GPa: from 0.7 to 1.0% for composition theoretical results were not taken into consideration beyond this limit due to the MBK, and from 0.6 to 0.9% for composition MBK+. Mass balance difficulties related to pl stabilization in pMELTS; the shape of the curve beyond calculations and SEM observations preclude the existence of a Ti- this limit is shown by thick dashed line for the source with 2000 ppm K2O. The open circles are the experimental data for composition MBK+ (930 ppm K2O) at rich accessory phase that would explain the different behavior of 1.3 GPa. The thin dashed lines subdivide the melt compositions into the following TiO2 at 1.3 GPa. fields: (I) phonolites; (II) tephriphonolites and trachyandesites; (III) phonotephrites The contrasted behavior between the series at 1 and 1.3 GPa and basaltic trachyandesites; (IV) trachybasalts; and (V) basalts (see Fig. 4 for the is not due to the effect of pressure because TiO is incompati- nomenclature of volcanic rocks). 2 ble in basaltic melts both at 1 and 1.5 GPa (Baker et al., 1995; Gaetani, 2004). The contrast presumably reflects in part a crystal- chyandesite melt for a source with 2000 ppm K2O (Fig. 6). Basalts = chemical control on element partitioning between minerals and sensu stricto only appear at F 4.3% for a K2O-free source and + F = 9.5% for a source with 2000 ppm K O. melt (Blundy et al., 1998), in relation to the enrichment in Na and 2 4+ Silica enrichment in low-degree melts of lherzolites has been other charge-balancing cations in cpx (the primary host for Ti ) observed in the experimental studies of Baker and Stolper (1994), near the solidus. We suggest, however, that the behavior of TiO2 in Baker et al. (1995), Hirschmann et al. (1998), and Robinson et al. our experiments is primarily controlled by major changes in melt (1998), as illustrated in Fig. 4. The liquid in equilibrium with an structure and polymerization at low degrees of melting (Gaetani, albite-bearing harzburgite residue at 1GPa is also strongly en- 2004; Huang et al., 2006). To illustrate this point, TiO2 in melt riched in silica, with ≈64% SiO2 and 12% Na2O (Falloon et al., is plotted as a function of melt polymerization (as embodied by 1997). Silica-rich liquids in equilibrium with ol + opx (±cpx, spl, NBO/T) in Fig. 7a, and as a function of the degree of silica under- pl) have been reported in partial melting experiments of Na2O- saturation (as embodied by the amount of normative nepheline or rich starting materials at 0.75 GPa (Chalot-Prat et al., 2010). In hyperstene) in Fig. 7b. On the right side of these two figures, TiO2 these systems, the enrichment in silica at low F is mainly a conse- increases with decreasing F (that is, from the right to the left) up quence of high concentrations in alkalis, which depolymerize the to a maximum of about 1%; on the left side, it decreases with de- liquid and lower the activity coefficient of SiO2. As the silica activ- creasing F . The maximum is located between NBO/T equal 0.3 and ity in peridotite melt is buffered by the equilibrium with magne- 0.6 in Fig. 7a, and at 10% nepheline in Fig. 7b. The decrease of TiO2 + sian ol opx, the decrease of the activity coefficient of SiO2 must at low F is mostly due to an increase of the partition coefficient of be compensated by an increase of the SiO2 concentration of the TiO2 between cpx and melt with increasing degree of melt poly- cpx/melt melt (Ryerson, 1985). This effect is supposed to be maximum at merization: D is >0.5 for NBO/T < 0.3, and up to 0.8–1.1 in P ≤ 1.5GPa(Hirschmann et al., 1998), but Davis and Hirschmann Ti phonolite melts (NBO/T < 0.2; Table 1), as compared to values in (2013) recently documented it to be significant at 3 GPa as well. the range 0.25–0.41 for melts with NBO/T > 0.5. Our results are in In addition to a strong enrichment in silica and alkalis, our low- agreement with those of Baker et al. (1995) who found an increase degree melts are also characterized by high Al2O3 contents (up to cpx/melt of D at low F , with the work of Gaetani (2004) who found 21%) and extreme depletions in MgO, FeO, and CaO (down to 2.7, Ti 2.9, and 2.1%, respectively; Fig. 3). The partial melt at the low- an increase of the compatibility of TiO2 and rare earth elements est degree of melting of fertile peridotite MM3 at 1 GPa (F ≈ 2%) in cpx with increasing melt polymerization near the peridotite is a trachyandesite with a silica enrichment close to the maxi- solidus, and with pMELTS calculations (Hirschmann et al., 1999). 4+ mum value reached in this study (see Fig. 4; Baker et al., 1995; By analogy with Ti , other highly charged cations and rare earth Hirschmann et al., 1998), but it is less enriched in Al2O3 and elements, which are also primarily hosted into cpx, are expected to alkalis, and still relatively rich in MgO, FeO, and CaO (5.7, 4.1, become more compatible near the peridotite solidus (Hirschmann and 6.5%, respectively). To our knowledge, degrees of depletion et al., 1999), especially in the presence of melts strongly enriched in MgO, FeO, and CaO as strong as in our study have not yet in silica and alkalis (Huang et al., 2006). Extraction of very low- been observed in experimental partial melts of a fertile peri- degree mantle melts with a high degree of polymerization should dotite at 1–1.5 GPa: near-solidus melts from the literature contain leave residues with specific trace element patterns. D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 327

◦ 1200 C (F = 4.4%). The pl composition in Falloon’s et al. experi- ments (An52Ab48 and An61Ab39 in runs T-4353 and T-4377, respec- tively) is close to the labradorite compositions in our run products (An54–57Ab41–43Or2–3), while pl in Kushiro’s experiment is more calcic (An72Ab28). The disappearance of pl is responsible for the inflexions of the compositional trends of the liquids observed for composition MBK at 1GPa (open square symbols in Fig. 2). The curves for Na2O and Al2O3 provide the best examples: they are first buffered at ≈6% and 19.5%, respectively, in the pl stability field (the three data points at F ≤ 3%), then, after pl disappearance, they show a continuous decrease with increasing F typical of an incompatible behavior.

5.4. Sources and geological settings required to generate low-degree mantle melts of phonolitic affinity

A specific feature of the K2O-enriched sources in our work is that they are either water free (in pMELTS calculations) or water poor (in the nominally anhydrous experiments). Spinel lherzolite xenoliths bearing alkali feldspar and containing 300 to 1000 ppm K2O, as described in Siberia (Ionov et al., 1995, 1999), provide natural examples of this type of sources. Alkali feldspar-bearing upper mantle peridotites are extremely rare, however: the typi- cal potassium carriers in peridotite xenoliths are phlogopite and/or amphibole (e.g., Menzies et al., 1987), indicating that large scale potassium enrichment of the subcontinental lithospheric mantle is almost invariably accompanied by the introduction of H2O. Low- degree melts in equilibrium with phlogopite (±amphibole) in hy- drous lherzolites at 1GPa contain 3–4% dissolved water, are en- riched in silica as our low-degree melts, but not as enriched in alkalis (Condamine and Médard, 2014): as a consequence, they are quartz normative and plot in the trachyte field of the TAS dia- gram. Thus our study is clearly not applicable to partial melting of mica-rich peridotites, in which the liquid may be in equilibrium with phlogopite over a large temperature interval. We emphasize, however, that melts in our low-F experiments (F = 1–2%) contain Fig. 7. (a) TiO2 in experimental partial melts versus the ratio of non-bridging oxy- ≈ = gens to tetrahedrally coordinated cation (NBO/T). Symbols are as follows: open 1% H2O (e.g., 1.1% H2O at F 1.7% in MBK3; see the Supple- squares, source MBK at 1 GPa; solid circles, same source at 1.3 GPa; open circles, mentary material) so that the bulk water content of the partially source MBK+ at 1.3 GPa (reversal experiment MBK+13 included). The thin dashed molten peridotite is 100 to 200 ppm. Accordingly, this work not line is the third order polynomial curve best fitting the whole data set. (b) TiO2 con- only applies to dry melting of a potassium-enriched mantle, but centrations in partial melts as a function of the percentage of normative nepheline also to the melting of lherzolites containing small quantities of or hyperstene (same symbols as in a). mica and/or amphibole (≈1% or less) above the breakdown tem- ◦ peratures of these hydrous phases (1050–1150 C; e.g. Conceição 5.3. Partial melting of fertile peridotites at 1 GPa: the role of plagioclase and Green, 2004; Condamine and Médard, 2014). The conclusion that low-degree melts may be phonolitic also applies to pyroxene- At 1GPa, the paragenesis of composition MBK is character- rich sources as long as some ol and opx are present (in addition to ized by the presence of both spinel and plagioclase in addition cpx) to maintain the silica content in the liquid at a high level. An to ol, opx, and cpx: the mass fraction of pl at the solidus tem- example is given in the Supplementary material (Section 3.4). ≈ ≈ perature is 3% and pl disappears at F 5%. We did not no- Given an appropriate source, the generation of a phonolite melt tice any nucleation difficulty for pl using a finely ground natu- requires that melting begins at P ≈ 1.5GPaor less (Supplementary ral spl lherzolite (devoid of pl) as starting material. The abun- Fig. S2a), and its preservation requires that the degree of melting dance of pl at 1GPa (Fig. 1c) and its absence at 1.3 GPa are remains low (as partial melts become basaltic at higher degrees of in agreement with an upper stability limit of plagioclase at the melting; Fig. 4). In an upwelling mantle with a potential temper- ◦ fertile peridotite solidus close to 1.3 GPa (Falloon et al., 1999; ature of 1280–1400 C (Herzberg et al., 2007), melting will begin Till et al., 2012), with the exact value depending on bulk CaO, at too high pressures to produce silica-enriched low-degree melts, Na2O, K2O, Cr2O3 and mg#. Comparison of our results with those especially if the mantle contains traces of water (Asimow et al., of Baker et al. (1995) and Borghini et al. (2010) illustrates the 2004), and if any phonolite melt is formed somehow at lower pres- importance of bulk composition on pl stability. Borghini et al. sures, it may get lost by mixing with non-phonolitic melts formed (2010) did not find pl above 0.8 GPa in a fertile peridotite con- underneath. Comparatively minor P –T perturbations may trigger taining less Na2O than MBK (0.26 vs. 0.35%). Similarly, Baker et al. mantle melting where the lithosphere is in extension and the (1995) did not find pl at the 1-GPa solidus of composition MM3, upper mantle has variable fertility and solidus temperature (e.g., which contains slightly less Na2O and more Cr2O3 than MBK (Ta- Anderson, 2005; Foulger, 2007): such a geological setting is clearly ble 1). This issue is controversial, however, because Falloon et al. more appropriate to the generation and preservation of phonolite (2008) found pl in a composition equivalent to MM3 at 1 GPa. melts. For instance, McKenzie (1989) proposed that a metasomatic Kushiro (1996) observed pl in lherzolite PHN1611 at 1GPa and layer enriched in K2O formed with time into the subcontinental 328 D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 lithospheric mantle by accumulation and freezing of small melt an initially uniform melt fraction. As the polycrystalline matrix is fractions from the asthenospheric mantle: remelting of this layer denser than the liquid, it compacts with time and the liquid ex- by adiabatic decompression related to extension of the lithosphere pelled by compaction forms a layer at the top of the system. For is a process that could yield large volumes of phonolite melt. given values of h and melt fraction, the melt viscosity, the den- sity contrast between melt and matrix, and the permeability of the 5.5. Direct generation of phonolites and nepheline syenites by partial matrix are the three major parameters controlling the efficiency melting of upper mantle rocks of melt extraction. In our calculations, we considered melt vis- cosities of the order of 102 Pa · s(Giordano et al., 2008), a melt Many felsic alkaline rocks (phonolites, nepheline syenites, etc.) density of 2500 kg/m3 (Seifert et al., 2013)and a matrix density are the products of crystal fractionation of basic precursors, such of 3200 kg/m3. Maumus et al. (2004) measured a dihedral angle ◦ as basanites or alkali basalts, at relatively shallow depths (Kyle et θ ≈ 50 for SiO2- and alkali-rich liquids close to phonolites in al., 1992; Ablay et al., 1998; Thompson et al., 2001). There are, contact with ol. In a partially molten rock with a dihedral angle ◦ however, a number of occurrences where these rocks cannot be θ<60 , the melt is interconnected at all melt fractions and the produced by low-pressure differentiation (Bailey, 1987). This is the permeability k at low melt fraction is proportional to the square case of phonolites and trachyandesites carrying ultramafic xeno- of melt fraction per unit volume, FV (von Bargen and Waff, 1986; liths of upper mantle origin (Wright, 1969; Green et al., 1974; McKenzie, 1989): Irving and Price, 1981; Dautria and Girod, 1983; Grant et al., 2013) 2 2 and of volcanic provinces with abundant phonolites but no or little a · F k = V associated basic to intermediate rocks (e.g., the plateau-type flood 3000 phonolites from the Kenya rift; Hay and Wendlandt, 1995). Mod- ≈ −3 els for a deep origin of felsic alkaline rocks include high-pressure where a is the grain size in the matrix (a 10 m). fractionation of a basanitic precursor at mantle depths (Irving and In a lherzolite source with 100 ppm K2O, the fraction of phono- = Price, 1981; Irving and Green, 2008), melting of alkali basalts un- lite melt is very small (0.27 wt.%, that is FV 0.0035; Fig. 6), and × −15 2 = derplated or injected into the lower crust (Hay and Wendlandt, so the permeability is only 4.0 10 m . Assuming h 1000 m, × 5 1995; Price et al., 2003), or low-degree melting of metasoma- it would take 3 10 years to extract a melt layer only 2.2 m thick tized mantle (Bailey, 1987; Riley et al., 1999). Dikes of silica- (see the Supplementary material for calculation details). By con- undersaturated felsic rocks intruding orogenic lherzolites have also trast, a source with 2000 ppm K2O yields almost 10 times more = been interpreted as low-degree mantle melts (Stähle et al., 1990; phonolite melt (2.6 wt.%, that is FV 0.033), and thus its perme- × −13 2 Pin et al., 2006). ability is two orders of magnitude larger: 3.6 10 m . As a The hypothesis that some phonolites or their plutonic equiva- result, for the same source thickness, the time scale of compaction × 4 lents (nepheline syenites) are produced directly by partial melting drops to 8 10 years and the layer of extracted melt is ten times of upper mantle rocks is strongly supported by our experimental thicker: 21 m. Accordingly, a moderate increase of the K2O con- results. Heldburg phonolite, Germany, is one of the best examples tent of mantle sources has a major effect both on the amount of of a phonolite carrying spinel lherzolite xenoliths (Irving and Price, phonolite melts produced and on the efficiency of melt separation 1981; Wedepohl et al., 1994; Grant et al., 2013). For many ele- (through the effect of melt fraction on permeability): the extrac- ments, there is a very good correspondence between partial melt tion of a body of primary mantle phonolite or nepheline syenite in run MBK+6 and Heldburg phonolite (Table 1). The correspon- sufficiently large to reach the upper crust or the Earth’s surface dence is not good, however, for MgO: with only 1.16% MgO, Held- presumably requires a large volume of K2O-enriched mantle. burg phonolite has a mg# of 51.8, which is too low for a primary mantle melt in equilibrium with MgO-rich olivine (mg# ≈90); for 5.6. Implications for mantle metasomatism comparison, the partial melt in run MBK+6 contains 2.72% MgO and has a mg# of 69.7. Accordingly, Heldburg phonolite cannot Metasomatism of the lithospheric mantle by the percolation be the direct equivalent of the high-mg# phonolite melts in our of small fractions of Si- and K-rich melts has been reported by experiments. It could be produced, however, by the separation of Bodinier et al. (1996) in spinel peridotites from the Ronda mas- 10–15% of ferromagnesian minerals from such high-mg# melts (at sif, Spain, and the East African Rift, Ethiopia, and by Pilet et al. upper mantle pressures to account for the incorporation of man- (2002) in the source of Cenozoic alkali basalts of the Cantal massif, tle xenoliths) or by partial melting of more pyroxene-rich mantle France. The migration of silica- and alkali-enriched liquids pro- veins with mg# <90. To test the first hypothesis, we performed a duced by low-degree melting of fertile peridotites, as in this study, ◦ crystallization experiment of phonolite gel +6 at 1 GPa-1150 C could be responsible for this type of metasomatism. The suites (CrysPho2; Table 1 and Supplementary Fig. S3): with less than of megacrysts observed in many alkaline basalt provinces (alkali 7% of crystallization (mostly cpx, plus minor ol), the MgO con- feldspar, clinopyroxene, etc.; Wright, 1969; Green et al., 1974; tent of the liquid dropped to 1.7% and its mg# to 60.3, but the Bailey, 1987) and sometimes associated with syenite xenoliths overall composition remains typically phonolitic and close to Held- (e.g., Upton et al., 1999) may be another manifestation of the circu- burg phonolite (Figs. 3–4). Therefore low-degree melting of a peri- lation and crystallization of Si- and K-rich melts in the lithospheric dotitic or more pyroxene-rich source (containing some water to mantle. account for the presence of phlogopite and amphibole phenocrysts) Detailed studies of mantle xenoliths worldwide have revealed at P ≤ 1.5GPais a plausible model for the formation of Held- the presence of alkali-rich glasses with SiO2 ≥ 55% either as inter- burg phonolite and other phonolites carrying peridotite xenoliths. granular patches, films and reaction rims (Draper and Green, 1997; Reevaluation of the geochemical data available on these rocks and Neumann and Wulff-Pedersen, 1997; Coltorti et al., 1999)or as acquisition of new isotopic data are required to establish unam- inclusions in primary minerals, notably olivines (e.g., Schiano and biguously a direct derivation by partial melting of a mantle source. Clochiatti, 1994). One interpretation of the intergranular glasses is We used the compaction model of McKenzie (1989) to estimate that they result from the reaction between a metasomatizing agent how fast a phonolite melt could be extracted from a lherzolite and primary mantle minerals (but this issue is still controversial; source by buoyancy (see Section 4 of the Supplementary mate- see Shaw et al., 2006, for a discussion). In particular, Coltorti et al. rial for details of the model). The lherzolite source is assumed to (2000) identified a Na- and a K-alkali silicate metasomatism that be a layer of thickness h with an interconnected melt network and they ascribed to the infiltration of basic melts, but the infiltration D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 329 of low-degree phonolitic melts of fertile peridotites may be an al- come less incompatible due to the abundance of pl, preventing a ternative explanation. strong enrichment of those two elements in the liquid.

5.7. Implications for the origin of glass inclusions in peridotite xenolith Acknowledgements minerals Jean-Luc Devidal is thanked for his efficient analytical assis- Silica- and alkali-rich glass inclusions showing similarities to tance with the electron microprobe and for the preparation of our experimental melts have been described in magnesian olivines synthetic gels. We are grateful to Mhammed Benbakkar and Jean- in spinel peridotite xenoliths from the sub-continental and sub- Marc Hénot for analytical assistance, to Christian Pin for discus- oceanic lithosphere (e.g., Schiano and Clocchiatti, 1994). They were sions on the occurrence of silica-undersaturated felsic rocks in interpreted as low-degree melts of spinel lherzolites formed at orogenic lherzolites, and to Pierre Condamine and Etienne Médard pressures around 1 GPa (Schiano et al., 1998) and acting as meta- for sharing their unpublished manuscript. The constant support of somatic agents, in good agreement with this study. We did not find Karine David to DL is gratefully acknowledged. We would like to olivine-hosted phonolitic glass inclusions in the literature data, but thank the anonymous reviewers for their constructive comments and John Brodholt for editing our manuscript. This work was initi- trachyandesitic to tephriphonolitic glass inclusions (54–57% SiO2; ated within the framework of the Research Training Network EU- 9–11% Na2O + K2O) are common. They presumably correspond to melt fractions too small to escape by buoyancy-driven porous flow ROMELT (HPRN-CT-2002-00211) of the European Community’s Hu- as discussed above, and that were ultimately trapped inside olivine man Potential Program. The experimental and analytical costs were grains as a result of grain boundary migration (Renner et al., 2002) covered by financial supports from the program SYSTER of the In- or thermally driven transcrystalline melt migration (Schiano et al., stitut National des Sciences de l’Univers (INSU-CNRS) and from the 2006). The software pMELTS could be used to compute the pres- Agence Nationale de la Recherche (ELECTROLITH project, contract sure and degree of melting from the composition of such glass no. ANR-2010-BLAN-621). SL’s work was supported by the National inclusions provided that the bulk xenolith composition is known. Science Foundation (EAR-1019886). The experimental facilities in Clermont-Ferrand benefited from the support of the French Gov- Some olivine-hosted glass inclusions in peridotite xenoliths ◦ ernment Laboratory of Excellence initiative n ANR-10-LABX-0006, contain SiO2 in excess of 60% and are trachytic in composi- the Région Auvergne and the European Regional Development tions (60–63% SiO2; 9–11% Na2O + K2O; Schiano et al., 1998; Schiano and Bourdon, 1999). Draper and Green (1997, 1999) also Fund. This is Laboratory of Excellence ClerVolc contribution num- found a class of mantle-xenolith glasses with broadly trachytic ber 111. compositions (Fig. 4), and they performed multiple saturation ex- Appendix A. Supplementary material periments that yielded melts with SiO2 ≈ 61% and Na2O + K2O ≈ 13% in equilibrium with a lherzolitic mineral assemblage at ◦ 1.2 GPa-1150 C. On the basis of these observations, it was one Supplementary material related to this article can be found on- of our initial objectives to produce trachytic compositions at low line at http://dx.doi.org/10.1016/j.epsl.2014.08.002. degrees of melting of a fertile peridotite. Our experiments and cal- culations show instead that low-degree melts of a dry lherzolite References at 1–1.3 GPa are phonolitic, with up to 14% Na2O + K2O but SiO2 ≤ 58%. We infer that the trachytic melt compositions in peridotite Ablay, G.J., Carroll, M.R., Palmer, M.R., Martí, J., Sparks, R.S.J., 1998. Basanite– phonolite lineages of the Teide-Pico Viejo volcanic complex, Tenerife, Canary xenoliths must be produced at pressures <1GPa and/or under Islands. J. Petrol. 39, 905–936. hydrous conditions. The association of a trachytic liquid with an Albarède, F., Provost, A., 1977. Petrologic and geochemical mass-balance equa- ol + opx + cpx assemblage at 1.2 GPa in Draper and Green’s tions: an algorithm for least-square fitting and general error analysis. Comput. (1999) study is presumably due to the effect of water. Accord- Geosci. 3, 309–326. Anderson, D.L., 2005. Scoring hotspots: the plume and plate paradigms. In: Foulger, ing to pMELTS, it is indeed necessary to add several hundred ppm ≈ G.R., et al. (Eds.), Plates, Plumes, and Paradigms. In: Spec. Pap., Geol. Soc. Am., of H2O into the source ( 2.5% H2O in the liquid) in order to keep vol. 388, pp. 31–54. opx stable in addition to ol and cpx in their experimental condi- Arevalo Jr., R., McDonough, W.F., Luong, M., 2009. The K/U ratio of the silicate Earth: tions. insights into mantle composition, structure and thermal evolution. Earth Planet. Sci. Lett. 278, 361–369. Asimow, P.D., Dixon, J.E., Langmuir, C.H., 2004. A hydrous melting and frac- 6. Conclusions tionation model for mid-ocean ridge basalts: applications to the Mid- Atlantic Ridge near the Azores. Geochem. Geophys. Geosyst. 5, Q01E16. Partial melts produced just above the solidus of fertile peri- http://dx.doi.org/10.1029/2003GC000568. Bailey, D.K., 1987. Mantle metasomatism – perspective and prospect. In: Fitton, J.G., dotites at 1–1.5 GPa have compositions very far from basalts and Upton, B.G.J. (Eds.), Alkaline Igneous Rocks. Geol. Soc. (Lond.) Spec. Publ. 30, classify as trachyandesites, tephriphonolites and even phonolites. 1–13. These silica- and alkali-rich melts have a high potential as metaso- Baker, M.B., Stolper, E.M., 1994. Determining the composition of high-pressure man- matic agents of the lithospheric mantle. Because their high degree tle melts using diamond aggregates. Geochim. Cosmochim. Acta 58, 2811–2827. Baker, M.B., Hirschmann, M.M., Ghiorso, M.S., Stolper, E.M., 1995. Compositions of of polymerization has important consequences for the partition- near-solidus peridotite melts from experiments and thermodynamic calcula- 4+ ing of minor and trace elements, as demonstrated here for Ti , tions. Nature 375, 308–311. they may lead to specific chemical and isotopic changes of the Blundy, J.D., Robinson, J.A.C., Wood, B.J., 1998. Heavy REE are compatible in clinopy- upper mantle. When a large volume of the upper mantle is en- roxene on the spinel lherzolite. Earth Planet. Sci. Lett. 160, 493–504. riched in potassium, the silica- and alkali-rich melts generated just Bodinier, J.-L., Merlet, C., Bedini, R.M., Simien, F., Remaidi, M., Garrido, C.J., 1996. Distribution of niobium, tantalum, and other highly incompatible trace elements above the peridotite solidus due to lithospheric extension could in the lithospheric mantle: the spinel paradox. Geochim. Cosmochim. Acta 60, be extracted and yield bodies of phonolite or nepheline syenite 545–550. sufficiently large to reach the upper crust or the Earth’s surface. Borghini, G., Fumagalli, P., Rampone, E., 2010. The stability of plagioclase in Production of phonolite liquids from a mantle source is presum- the upper mantle: subsolidus experiments on fertile and depleted lherzolite. J. Petrol. 51, 229–254. ably restricted to a narrow range of pressures (between ≈1.5 and Chalot-Prat, F., Falloon, T.J., Green, D.H., Hibberson, W.O., 2010. An experimental 1GPa): strong SiO2 enrichment of low-degree melts is not ex- study of liquid compositions in equilibrium with plagioclase + spinel lherzo- pected above ≈1.5 GPa, and at 1GPa or less, Na2O and K2O be- lite at low pressures (0.75 GPa). J. Petrol. 51, 2349–2376. 330 D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331

Coltorti, M., Bonadiman, C., Hinton, R.W., Siena, F., Upton, B.G.J., 1999. Carbonatite Irving, A.J., Price, R.C., 1981. Geochemistry and evolution of lherzolite-bearing metasomatism of the oceanic upper mantle: evidence from clinopyroxenes and phonolitic lavas from Nigeria, Australia, East Germany and New Zealand. glasses in ultramafic xenoliths of Grande Comore, Indian Ocean. J. Petrol. 40, Geochim. Cosmochim. Acta 45, 1309–1320. 133–165. Kushiro, I., 1996. Partial melting of a fertile mantle peridotite at high pressures: an Coltorti, M., Beccaluva, L., Bonadiman, C., Salvini, L., Siena, F., 2000. Glasses in man- experimental study using aggregates of diamond. In: Basu, A., Hart, S. (Eds.), tle xenoliths as geochemical indicators of metasomatic agents. Earth Planet. Sci. Earth Processes: Reading the Isotopic Code. In: Geophys. Monogr., vol. 95. Amer. Lett. 183, 303–320. Geophys. Union, Washington, DC, pp. 109–122. Conceição, R.V., Green, D.H., 2004. Derivation of potassic (shoshonitic) magmas by Kyle, P.R., Moore, J.A., Thirlwall, M.F., 1992. Petrologic evolution of anorthoclase decompression melting of phlogopite + pargasite lherzolite. Lithos 72, 209–229. phonolite lavas at Mount Erebus, Ross Island, Antarctica. J. Petrol. 33, 849–875. Condamine, P., Médard, E., 2014. Experimental melting of phlogopite-bearing man- Lambart, S., Laporte, D., Schiano, P., 2009. An experimental study of focused magma tle at 1 GPa: implications for potassic magmatism. Earth Planet. Sci. Lett. 397, transport and basalt–peridotite interactions beneath mid-ocean ridges: impli- 80–92. cations for the generation of primitive MORB compositions. Contrib. Mineral. Dautria, J.M., Girod, M., 1983. The upper mantle beneath eastern Nigeria: inferences Petrol. 157, 429–451. from ultramafic xenoliths in Jos and Biu volcanics. J. Afr. Earth Sci. 1, 331–338. Lambart, S., Laporte, D., Schiano, P., 2013. Markers of the pyroxenite contribution in Davis, F.A., Hirschmann, M.M., 2013. The effects of K2O on the compositions of the major-element compositions of oceanic basalts: review of the experimental near-solidus melts of garnet peridotite at 3 GPa and the origin of basalts from constraints. Lithos 160–161, 14–36. enriched mantle. Contrib. Mineral. Petrol. 166, 1029–1046. Laporte, D., Toplis, M., Seyler, M., Devidal, J.-L., 2004. A new experimental technique Draper, D.S., Green, T.H., 1997. P –T phase relations of silicic, alkaline, aluminous for extracting liquids from peridotite at very low degrees of melting: application mantle–xenolith glasses under anhydrous and C–O–H fluid saturated conditions. to partial melting of depleted peridotite. Contrib. Mineral. Petrol. 146, 463–484. J. Petrol. 38, 1187–1224. Le Bas, M.J., Le Maitre, R.W., Streckeisen, A., Zanettin, B., 1986. A chemical classi- Draper, D.S., Green, T.H., 1999. P –T phase relations of silicic, alkaline, aluminous liq- fication of volcanic rocks based on the total alkali-silica diagram. J. Petrol. 27, uids: new results and applications to mantle melting and metasomatism. Earth 745–750. Planet. Sci. Lett. 170, 255–268. Maumus, J., Laporte, D., Schiano, P., 2004. Dihedral angle measurements and infil- Falloon, T.J., Green, D.H., O’Neill, H.St.C., Hibberson, W.O., 1997. Experimental tests tration property of SiO2-rich melts in mantle peridotite assemblages. Contrib. of low degree peridotite partial melt compositions: implications for the nature Mineral. Petrol. 148, 1–12. of anhydrous near-solidus peridotite melts at 1 GPa. Earth Planet. Sci. Lett. 152, McDonough, W.F., Sun, S.S., 1995. The composition of the Earth. Chem. Geol. 120, 149–162. 223–253. Falloon, T.J., Green, D.H., Danyushevsky, L.V., Faul, U.H., 1999. Peridotite melting at McKenzie, D., 1989. Some remarks on the movement of small melt fractions in the 1.0 and 1.5 GPa: an experimental evaluation of techniques using diamond aggre- mantle. Earth Planet. Sci. Lett. 95, 53–72. gates and mineral mixes for determination of near-solidus melts. J. Petrol. 40, Menzies, M.A., Rogers, N., Tindle, A., Hawkesworth, C.J., 1987. Metasomatic and 1343–1375. enrichment processes in lithospheric peridotites, an effect of asthenosphere– Falloon, T.J., Green, D.H., Danyushevsky, L.V., McNeill, A.W., 2008. The composition lithosphere interaction. In: Menzies, M.A., Hawkesworth, C.J. (Eds.), Mantle of near-solidus partial melts of fertile peridotite at 1 and 1.5 GPa: implications Metasomatism. Academic Press, pp. 313–361. for the petrogenesis of MORB. J. Petrol. 49, 591–613. Mysen, B.O., Richet, P., 2005. Melt and glass structure. Basic concepts. In: Mysen, Foulger, G.R., 2007. The “plate” model for the genesis of melting anomalies. In: Foul- B.O., Richet, P. (Eds.), Silicate Glasses and Melts: Properties and Structure. In: ger, G.R., Jurdy, D.M. (Eds.), Plates, Plumes, and Planetary Processes. In: Spec. Developments in Geochemistry, vol. 10, pp. 101–129. Pap., Geol. Soc. Am., vol. 430, pp. 1–28. Neumann, E.-R., Wulff-Pedersen, E., 1997. The origin of highly silicic glass in mantle Gaetani, G.A., 2004. The influence of melt structure on trace-element partitioning xenoliths from the Canary Islands. J. Petrol. 38, 1513–1539. near the peridotite solidus. Contrib. Mineral. Petrol. 147, 511–527. Ottolini, L., Hawthorne, F.C., 2001. SIMS ionization of hydrogen in silicates: acase Ghiorso, M.S., Hirschmann, M.M., Reiners, P.W., Kress III, V.C., 2002. The pMELTS: study of kornerupine. J. Anal. At. Spectrom. 6, 1266–1270. a revision of MELTS for improved calculation of phase relations and major ele- Ottolini, L., Bottazzi, P., Zanetti, A., Vannucci, R., 1995. Determination of hydrogen in ment partitioning related to partial melting of the mantle to 3 GPa. Geochem. silicates by secondary ion mass spectrometry. Analyst 120, 1309–1313. Geophys. Geosyst. 3 (5). http://dx.doi.org/10.1029/2001GC000217. Ottolini, L., Cámara, F., Hawthorne, F.C., Stirling, J., 2002. SIMS matrix effects in the Giordano, D., Russell, J.K., Dingwell, D.B., 2008. Viscosity of magmatic liquids: analysis of light elements in silicate minerals: comparison with SREF and EMPA amodel. Earth Planet. Sci. Lett. 271, 123–134. data. Am. Mineral. 87, 1477–1485. Grant, T.B., Milke, R., Pandey, S., Jahnke, H., 2013. The Heldburg phonolite, Central Pickering-Witter, J., Johnston, A.D., 2000. The effects of variable bulk composition Germany: reactions between phonolite and xenocrysts from the upper mantle on the melting systematics of fertile peridotitic assemblages. Contrib. Mineral. and lower crust. Lithos 182–183, 86–101. Petrol. 140, 190–211. Green, D.H., Edgar, A.D., Beasley, P., Kiss, E., Ware, N.G., 1974. Upper mantle source Pilet, S., Hernandez, J., Villemant, B., 2002. Evidence for high silicic melt circulation for some hawaiites, mugearites and benmoreites. Contrib. Mineral. Petrol. 48, and metasomatic events in the mantle beneath alkaline provinces: the Na–Fe- 33–43. augitic green-core pyroxenes in the Tertiary alkali basalts of the Cantal massif Hay, D.E., Wendlandt, R.F., 1995. The origin of Kenya rift plateau-type flood phono- (French Massif Central). Mineral. Petrol. 76, 39–62. lites: results of high-pressure/high-temperature experiments in the systems Pin, C., Monchoux, P., Paquette, J.-L., Azambre, B., Wang, Ru Cheng, Martin, R.F., 2006. phonolite–H2O and phonolite–H2O–CO2. J. Geophys. Res. 100, 401–410. Igneous albitite dikes in orogenic lherzolites, Western Pyrénées, France: a pos- Herzberg, C., Asimow, P.D., Arndt, N., Niu, Y., Lesher, C.M., Fitton, J.G., Cheadle, M.J., sible source for corundum and alkali feldspar xenocrysts in basaltic terranes. II. Saunders, A.D., 2007. Temperatures in ambient mantle and plumes: constraints Geochemical and petrogenetic considerations. Can. Mineral. 44, 843–856. from basalts, picrites, and komatiites. Geochem. Geophys. Geosyst. 8, Q02006. Presnall, D.C., Gudfinnsson, G.H., Walter, M.J., 2002. Generation of mid-ocean ridge http://dx.doi.org/10.1029/2006GC001390. basalts at pressures from 1 to 7 GPa. Geochim. Cosmochim. Acta 66, 2073–2090. Hirose, K., Kushiro, I., 1993. Partial melting of dry peridotites at high pressures: Price, R.C., Cooper, A.F., Woodhead, J.D., Cartwright, I., 2003. Phonolitic dia- determination of compositions of melts segregated from peridotite using aggre- tremes within the Dunedin volcano, South Island, New Zealand. J. Petrol. 44, gates of diamond. Earth Planet. Sci. Lett. 114, 447–489. 2053–2080. Hirschmann, M.M., Baker, M.B., Stolper, E.M., 1998. The effect of alkalis on the silica Renner, J., Evans, B., Hirth, G., 2002. Grain growth and inclusion formation in par- content of mantle-derived melts. Geochim. Cosmochim. Acta 62, 883–902. tially molten carbonate rocks. Contrib. Mineral. Petrol. 142, 501–514. Hirschmann, M.M., Ghiorso, M.S., Stolper, E.M., 1999. Calculation of peridotite par- Riley, T.R., Bailey, D.K., Harmer, R.E., Liebsch, H., Lloyd, F.E., Palmer, M.R., 1999. tial melting from thermodynamic models of minerals and melts. II. Isobaric Isotopic and geochemical investigation of a carbonatite–syenite–phonolite dia- variations in melts near the solidus and owing to variable source composition. treme, West Eifel (Germany). Mineral. Mag. 63, 615–631. J. Petrol. 40, 297–313. Robinson, J.A.C., Wood, B.J., 1998. The depth of the spinel to garnet transition at the Huang, F., Lundstrom, C.C., McDonough, W.F., 2006. Effect of melt structure on peridotite solidus. Earth Planet. Sci. Lett. 164, 277–284. trace-element partitioning between clinopyroxene and silicic, alkaline, alumi- Robinson, J.A.C., Wood, B.J., Blundy, J.D., 1998. The beginning of melting of fertile nous melts. Am. Mineral. 91, 1385–1400. and depleted peridotite at 1.5 GPa. Earth Planet. Sci. Lett. 155, 97–111. Ionov, D.A., O’Reilly, S.Y., Ashchepkov, I.V., 1995. Feldspar-bearing lherzolite xeno- Ryerson, F.J., 1985. Oxide solution mechanisms in silicate melts: systematic varia- liths in alkali basalts from Hamar-Daban, southern Baikal region, Russia. Contrib. tions in the activity coefficient of SiO2. Geochim. Cosmochim. Acta 49, 637–649. Mineral. Petrol. 122, 174–190. Schiano, P., Bourdon, B., 1999. On the preservation of mantle information in ultra- Ionov, D.A., Grégoire, M., Prikhod’ko, V.S., 1999. Feldspar–Ti-oxide metasomatism in mafic nodules: glass inclusions within minerals versus interstitial glasses. Earth off-cratonic continental and oceanic upper mantle. Earth Planet. Sci. Lett. 165, Planet. Sci. Lett. 169, 173–188. 37–44. Schiano, P., Clocchiatti, R., 1994. Worldwide occurrence of silica-rich melts in sub- Irving, A.J., Green, D.H., 2008. Phase relationships of hydrous alkalic magmas at continental and sub-oceanic mantle minerals. Nature 368, 621–624. high pressures: production of nepheline hawaiitic to mugearitic liquids by Schiano, P., Bourdon, B., Clocchiatti, R., Massare, D., Varela, M.E., Bottinga, Y., 1998. amphibole-dominated fractional crystallization within the lithospheric mantle. Low-degree partial melting trends recorded in upper mantle minerals. Earth J. Petrol. 49, 741–756. Planet. Sci. Lett. 160, 537–550. D. Laporte et al. / Earth and Planetary Science Letters 404 (2014) 319–331 331

Schiano, P., Provost, A., Clocchiatti, R., Faure, F., 2006. Transcrystalline melt migration Till, C.B., Grove, T.L., Krawczynski, M.J., 2012. A melting model for variably depleted and Earth’s mantle. Science 314, 970–974. and enriched lherzolite in the plagioclase and spinel stability fields. J. Geophys. Scordari, F., Dyar, M.D., Schingaro, E., Lacalamita, M., Ottolini, L., 2010. XRD, micro- Res. 117, B06206. Amer. Geophys. Union. XANES, EMPA, and SIMS investigation on phlogopite single crystals from Mt, Upton, B.G.J., Hinton, R.W., Aspen, P., Finch, A., Valley, J.W., 1999. Megacrysts and Vulture (Italy). Am. Mineral. 95, 1657–1670. associated xenoliths: evidence for migration of geochemically enriched melts in Seifert, R., Malfait, W.J., Petitgirard, S., Sanchez-Valle, C., 2013. Density of phono- the upper mantle beneath Scotland. J. Petrol. 40, 935–956. litic magmas and time scales of crystal fractionation in magma chambers. Earth von Bargen, N., Waff, H.S., 1986. Permeabilities, interfacial areas and curvatures of Planet. Sci. Lett. 381, 12–20. partially molten systems: results of numerical computations of equilibrium mi- Shaw, C.S.J., Heidelbach, F., Dingwell, D.B., 2006. The origin of reaction textures in crostructures. J. Geophys. Res. 91, 9261–9276. mantle peridotite xenoliths from Sal Island, Cape Verde: the case for “metaso- matism” by the host lava. Contrib. Mineral. Petrol. 151, 681–697. Walter, M.J., Sisson, T.W., Presnall, D.C., 1995. A mass proportion method for cal- Smith, P.M., Asimow, P.D., 2005. Adiabat_1ph: anew public front-end to the culating melting reactions and application to melting of model upper mantle MELTS, pMELTS, and pHMELTS models. Geochem. Geophys. Geosyst. 6 (Q02004). lherzolite. Earth Planet. Sci. Lett. 135, 77–90. http://dx.doi.org/10.1029/2004GC000816. Wasylenki, L.E., Baker, M.B., Kent, A.J.R., Stolper, E.M., 2003. Near-solidus melting of Stähle, V., Frenzel, G., Kober, B., Michard, A., Puchelt, H., Schneider, W., 1990. Zircon the shallow upper mantle: partial melting experiments on depleted peridotite. syenite pegmatites in the Finero peridotite (Ivrea zone): evidence for a syenite J. Petrol. 44, 1163–1191. from a mantle source. Earth Planet. Sci. Lett. 101, 196–205. Wedepohl, K.H., Gohn, E., Hartmann, G., 1994. Cenozoic alkali basaltic magmas Takahashi, E., Kushiro, I., 1983. Melting of a dry peridotite at high pressures and of western Germany and their products of differentiation. Contrib. Mineral. basalt magma genesis. Am. Mineral. 68, 859–879. Petrol. 115, 253–278. Taylor, S.R., McLennan, S.M., 1985. The Continental Crust: Its Composition and Evo- lution. Blackwell Scientific Publications. 312 pp. Workman, R.K., Hart, S.R., 2005. Major and trace element composition of the de- Thompson, G.M., Smith, I.E.M., Malpas, J.G., 2001. Origin of oceanic phonolites by pleted MORB mantle (DMM). Earth Planet. Sci. Lett. 231, 53–72. crystal fractionation and the problem of the Daly gap: an example from Raro- Wright, J.B., 1969. Olivine nodules and related inclusions in trachyte from the Jos tonga. Contrib. Mineral. Petrol. 142, 336–346. Plateau, Nigeria. Mineral. Mag. 37, 370–374.